Principles of Digital Image Synthesis

Rating:
Summary: Great Book
Review: Among all of the CG books on my shelf, this is the only one which binds the CG subjects to physics origin so deeply. Glassner prepared a very nice collection of reference information, explained the historical reasons of several confusing stuff in CG.

It's true that it has number of mistakes / typos but there's an online errata .... Once you check and note down the errata in the proper places of the book - which may take your 1hr at the most-, nothing will remain to complain about this book.

If you are serious about CG, you'll love the information in this book. It's a bit expensive but surely worth the price.

Rating:
Summary: Graphics is Math, Physics, Perception, & Computation
Review: Back in the old days, Computer Graphics was a big bag of tricks for making cool images. To make the pictures look better and better, the research community stumbled into areas we didn't originally know were important. We now see that clipping, viewports, line-drawing, and specular lights are not so fundamental. Instead, we're understanding that items in the bag of graphics tricks were often shortcuts to solving an integral equation for heat transfer. Who would have thought it?

This textbook is the first comprehensive treatment of Computer Graphics to convey the deeper understanding that researchers have finally begun to make peace with. It's not always easy. That marginal lecture on de-aliasing in your graphics class? It turns out to be hugely significant. Sampling and reconstruction pervade graphics algorithms, and the first 10 chapters cover the topic extensively. That reflectance distribution function you saw at the end of the semester? It's not an advanced topic. It's what realistic rendering is built from. How to represent it, evaluate it, and integrate it are the concerns of the next 10 chapters.

The hypothetical Ideal Graphics Professional has majored in Math, Computer Science, Physics, Perceptual Psychology, and Mechanical Engineering. No one has that background, but if you majored in any of these subjects and then patiently read this book, you will appreciate how the themes combine in a remarkable way whenever a pixel is drawn.

Rating:
Summary: Disappointed
Review: I was expecting much more out of this 2 volume set. The books are full of mistakes, especially in formulas. If you plan on purchasing these books, make sure to download and print out the errata as well. It could save many headaches in trying to understand formulas that don't agree with the accompanying explanation. I don't doubt that Glassner is a very intelligent man, but his descriptions are somewhat dense and difficult to decypher sometimes. My recommendation: get an ACM SIGGRAPH membership to get access to many of the papers in this field and get the algorithms straight from the source.

Rating:
Summary: Disappointed
Review: I was quite disappointed with the number of mistakes in this book as I would have liked to have used it as a reference.

Many sections were grabbed from research papers which I am familiar with and no attempt was made to keep notations consistent between papers. This meant that as you are reading the notations will suddenly switch on you.

I think the undertaking was too ambitious although the topics are quite pertinent.

Perhaps the next release will address these problems.

Rating:
Summary: Strong, serious work on the science behind computer graphics
Review: This book is the first serious work to bring together elements from disparate areas of science, such as optics, psychology, and monte carlo sampling, and show how these are used in the field of computer graphics. It is not a book about how to code various algorithms, rather it is a place to go to learn about what has been done in the field and to gain inspiration for further research. It is for the most part readable, but not light reading. These two volumes are a great time saver if you need to understand the relevant theory in a particular area of research; the author has combed through literally thousands of papers and books and presented the essence of many of the best of them. Its high rating can be justified by the fact that it is the only work of its kind; that it is well illustrated, informative, and useful makes it all the better.

Rating:
Summary: A unique resource
Review: This is the one really fundamental book on rendering computer images. Watt's book is an excellent introduction to many basic principles, but Glassner's volumes are an advanced text suitable even for experienced graphics professionals. Here you can find a description of fourier analysis and wavelets using the same notation, a survey of appearance science, and a good description of the physics underlying rendering algorithms. It really has everything. There are some typos in some of the equations, but the errata are available online. I find myself using these volumes all the time.

Rating:
Summary: A Unique & Authorative Resource
Review: This set of books is a unique & authorative resource guiding you though a whole range of 2D and 3D graphics concepts and algorithms. I have referred to it many times for explainations and definitions related to 3D rendering and shading models, among other things. However, this is NOT just an encyclopedia or compilation of other people's works - the author (a respected senior CG researcher himself) has gone through a whole body of knowledge, explaining and linking each concept, and including the CG algorithms in a broader context along with concepts of human perception and underlying theories. Nothing like this has ever been written - I would recommend it to anyone who can understand it, with the sole warning that some college-level math (at least a little calculus) is needed in order to read and follow much of what he covers.

Rating:
Summary: Excellent
Review: Volume 1:
This book is comprehensive in scope and one of the most well-written technical books in existence. In the preface the author states 'I love to write', and considering the exceptional quality of this book, this indeed shows through.

The first part of the book covers the human visual system, the understanding of which is fundamental to designing effective computer graphics. Several interesting topics are discussed, including Mach bands, color opponency, perceptual color matching, MacAdam ellipses, RGB color space, and gamut mapping.

The second part covers more technical matters, namely that of signal processing. The mathematical background assumed of the reader increases dramatically in this part; some exposure to elementary calculus and differential equations would suffice. The author does a good job of explaining such concepts as linear operators and the Dirac bracket notation. The pictorial representation he gives of the convolution operation is very helpful. In addition, Fourier analysis is presented at a level that makes it very clear exactly what is happening to signals, both discrete and continuous, when taking the Fourier transform. The Fast Fourier transform is not discussed however, dissapointingly. Suprisingly, a whole chapter is devoted to wavelet transforms, a topic usually not included at this level. Wavelets are used as a tool to deal with nonstationary signals. Usually discussed at a very abstract level, the presentation here is crystal clear and vey intutive, and the reader will take away a deeper appreciation of these objects than what could have been obtained from the usual presentations.

Chapter 7 is one of the most important in the book for it covers Monte Carlo techniques for evaluating the integrals that arise in image processing. The speed of convergance of Monte Carlo is addressed, along with how to estimate confidence levels when the parent distribution is normal. The author presents five different ways of doing 'blind' Monte Carlo, including rejection, blind stratified, weighted, and quasi Monte Carlo. Quasi Monte Carlo has taken on particular importance in recent years wherever Monte Carlo techniques are used. The author also presents four different ways of doing 'informed' Monte Carlo, i.e. when some information about the signal is known.

Uniform sampling of continuous signals is done in the next chapter. After discussing an example of sampling and reconstruction, the author outlines in detail the mathematical theory behind the uniform sampling and reconstruction of one-and two-dimensional signals. The chapter ends with a discussion of a technique to reduce aliasing artifacts called supersampling.

The next chapter covers nonuniform sampling and reconstruction. Naturally this is more complicated from a mathematical standpoint, due to the role of stochastic processes, but the author does a good job of discussing the relevant concepts. Most interesting is his treatment of the duality between aliasing and noise.

Chapter 10 surveys some of the more modern and practical techniques used for sampling and reconstruction of two-dimensional signals. Uniform sampling is discussed in terms of rectangular and hexagonal lattices; nonuniform sampling in terms of Poisson sampling and N-books sampling. Pseudocode is given for the decreasing radius algorithm. The concept of a refinement test is introduced and broken down into five categories, each of which is discussed in detail. The refinement test allows one to decide when more samples are needed in a neighborhood, and refinement geometry indicates where the samples are to be placed. Refinement geometry is discussed in this chapter also, with linear and area bisection techniques outlined, along with multiple-level and tree-based sampling. Techniques for interpolation and reconstruction, such as warping are also treated, and the author gives brief overviews of one-dimensional and two-dimensional sampling theorems. Numerous other methods, going by several different names are also discussed.

A very large set of references is given at the end of the book, covering a wide variety of topics in computer graphics and mathematical formalism. I have not read the second volume, but I am sure it respects the high quality of the first.